Limits and errors

Handle Enterprise API limits, HTTP failures, and realtime session errors.

Default limits may be adjusted for an approved organization during onboarding.

Limit	Default
Batch requests	60 requests per minute per account
Batch audio upload	100 MB
Concurrent realtime streams	5 per account
Realtime session duration	10 minutes
Realtime context	128 KB

Batch errors

Errors include a machine-readable code and human-readable error message.

{
  "error": "Request body too large",
  "code": "REQUEST_TOO_LARGE"
}

Status	Example code	Handling
`400`	`BAD_REQUEST`	Correct the request; do not retry unchanged.
`401`	Authentication error	Supply exactly one valid credential.
`403`	Origin not allowed	Correct the origin or token binding.
`408`	`REQUEST_CANCELLED`	Retry with bounded exponential backoff.
`413`	`REQUEST_TOO_LARGE`	Reduce the upload size.
`429`	Rate limited	Wait for `Retry-After`, then retry.
`500`	`PROCESSING_FAILED`	Do not retry indefinitely; contact support if persistent.
`503`	`TRANSCRIPTION_TIMEOUT` or `TRANSCRIPTION_UNAVAILABLE`	Retry with backoff and jitter.
`504`	`UPSTREAM_TIMEOUT`	Retry with backoff and jitter.

The X-Dictate-Attempts response header reports how many processing attempts were used. Preserve transcription_id across a retry so Monologue can correlate the requests.

Realtime errors

Realtime failures arrive as session.error:

{
  "type": "session.error",
  "code": "...",
  "message": "...",
  "retry_via_async": true
}

When retry_via_async is true, close the stream and submit the buffered audio to the batch endpoint. Do not automatically replay audio unless your product can prevent duplicate output and the user still expects the result.